Abstract
Statistical machine translation (SMT) refers to using probabilistic methods of learning translation process primarily from the parallel text. In SMT, the linguistic information such as morphology and syntax can be added to the parallel text for improved results. However, adding such linguistic matter is costly, in terms of time and expert effort. Here, we introduce a technique that can learn better shapes (morphological process) and more appropriate positioning (syntactic realization) of target words, without linguistic annotations. Our method improves result iteratively over multiple passes of translation. Our experiments showed better accuracy of translation, using a well-known scoring tool. There is no language specific step in this technique.
Highlights
Recent trend in machine translation is mostly towards datadriven methods including Statistical Machine Translation (SMT), which uses parallel text
A freely available toolkit for training and decoding of SMT systems, Moses [3] is used in our experiments, along with the supportive tools [4] for intermediate tasks like text alignment
Open source tools [5] are used for English, and locally developed morphology analyzer [6] and POS tagger [7] of Urdu are used for morphosyntactic experiment
Summary
Recent trend in machine translation is mostly towards datadriven methods including Statistical Machine Translation (SMT), which uses parallel text. This approach learns translation through phrase alignments [1] which are based on word alignments. In SMT, the morphological information improves learnability for realizing the correct shape of words, especially for morphologically rich languages like Arabic and Urdu. The experiments for baseline and proposed technique, both, use plain parallel text. Improved the shapes and arrangements of words on the target side by using the SMT process iteratively, to incrementally learn such information from the parallel text itself.
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
